Agent-based Interfaces u u u

Software Agents act on behalf of people in the electronic world. Agents can perform repetitive tasks E.g. Web crawlers and filters

Agents can learn from user’s actions Difficult to find a language suitable for user and agent User may not receive immediate feedback

Apple HyperCard: earliest example of AI Agent learning based on user actions. Eager had a clear ‘embodiment’, the cat on screen User free to ignore agent

MS Excel, intelligent functions but diffused, not embodied. Office assistant

Agent-based systems include aspects of both language and action paradigms. Old command based systems acted as intermediaries. Agents act on the user’s behalf

Direct manipulation emphasises the user’s own actions. The Agent is usually acting in an environment the user can also act upon.

The End